🇪🇸 Español 🇬🇧 English 🇧🇷 Português

Native tools (I): close conversation, assign to person or team, and add a note

The three foundational tools every AI agent in Omnifox has: closing the conversation, handing it off to a human, and leaving an internal note.

Jul 11, 2026

These three tools are the foundation of any AI agent in Omnifox and are available from the entry AI-enabled plan — they're the first ones you should turn on for a new agent.

Close conversation (close_conversation)

Marks the conversation as closed once the customer confirms their issue is resolved.

  • When to enable it: the customer confirmed the problem is solved, said goodbye ("thanks", "all set"), and there's nothing left open.
  • When NOT to: the customer still has questions, is unhappy, or you owe them something (info, a callback, etc.).
  • Effect: the conversation moves to "closed" status and stops counting as open in metrics and the inbox.

Assign to person or team (assign_to_user_or_team)

Hands the conversation off to a specific human or an entire team.

  • Parameters: target_type ("user" or "team"), target_name (team/person name, recommended) or target_id (numeric), and an optional reason.
  • You can reference by name ("Sales team", "John Smith") without memorizing IDs — the system resolves the lookup automatically. If there are multiple matches, the bot gets the candidates and can ask the customer to clarify.
  • Typical uses: routing by category (sales, support, billing), escalating complaints, or handing off to a human who already knows the case.

Add internal note (add_comment)

Leaves a note visible ONLY to your team — the customer never sees it.

  • Parameter: comment (text, max 2000 characters).
  • When to use it: before escalating (case summary), when capturing sensitive data (ID number, tax ID), when there's a payment promise, or to flag signals (upset customer, churn risk, upsell opportunity).

Configuration

Each tool is toggled independently per agent, under the editor's Custom actions tab. Turning a tool on isn't enough by itself — the agent's system prompt has to explicitly tell it WHEN to use it.

Example

Sample prompt for a support agent:

"If the customer confirms their issue is resolved and says goodbye, close the conversation with !close_conversation. If the case is technical and beyond your knowledge, first leave an !add_comment with a summary and the steps already tried, then !assign_to_user_or_team to the Support team."

Tips

  • Combine all three: leaving the internal note BEFORE handing off saves the human picking up the case a lot of time.
  • Test the full flow in the Sandbox before publishing it, checking the trace of each tool call.

Troubleshooting

  • If the agent never closes or hands off conversations, first check that the tool is enabled under Custom actions, then confirm the prompt describes the trigger condition (see the AI agent troubleshooting article).
Was this helpful?

Related articles